DoE design

#1
I would like to design a chemical process that has a raw material (A), a supplement (B), and a catalyst (C).
I have to screen the best raw material (from 15 raw material types), the best supplement (from 8 supplements) and the best catalyst ( from 4 catalysts) based on DoE. Please suggest a DoE type where I can get a process maximum yield (response) from the right combination of raw material, supplement and catalyst.
 

Miner

TS Contributor
#2
Ouch! I hope you have a lot of time and money for this. This type of experiment will require a General Full Factorial design. The minimum number of experiments required is 480 and that is for an unreplicated design. There is no way that you can run a fractionated design. Is there any non-experimental way (i.e., engineering judgment) to reduce the list to a smaller number, say 4 each? That would reduce the number of experiments to 64. I know chemical processes also have a large number of 2 and 3-way interactions.

One area of concern I have is that you would have to run this at a fixed mixture ratio. Potentially you could find that catalyst B works best at that ratio. However, catalyst D might have performed even better at a higher ratio.
 
#3
Maybe it turns out that the 15 raw material consists of factors like a 2*2*3 + 3 additional, så that the 15 are created partly as a factorial. Then it can be formulated as a fractional factorial.

Tell us more.
 
#4
Ouch! I hope you have a lot of time and money for this. This type of experiment will require a General Full Factorial design. The minimum number of experiments required is 480 and that is for an unreplicated design. There is no way that you can run a fractionated design. Is there any non-experimental way (i.e., engineering judgment) to reduce the list to a smaller number, say 4 each? That would reduce the number of experiments to 64. I know chemical processes also have a large number of 2 and 3-way interactions.

One area of concern I have is that you would have to run this at a fixed mixture ratio. Potentially you could find that catalyst B works best at that ratio. However, catalyst D might have performed even better at a higher ratio.
Thanking you for the reply.
480 runs is really a big deal; I would prefer OFAT over 480 runs but OFAT has some limitations like 1) I have to wait for the results of one factor studied before I can start studying the 2nd factor; that's time consuming. 2) I would miss out important interactions between three factors.

I guess splitting the design into 3-4 smaller DoEs is a good idea. But there should be a linkage or final DoE where I can relate smaller designs, may be RSM finally.

For interactions, I would be interested in two way interactions only. Higher order interactions would not be important for initial screening.

Generally in a process, a single raw material and a supplement are added in a definite percentage depending on the type of supplement. Supplement-1 has to be added at 10% concentration (i.e., raw material is 90%), Supplement-2 at 20%, Supplement-3 at 15% and supplement-4 at 25%. However, catalyst is added at a constant concentration that's independent of raw material and supplement concentration (negligible). I'm not sure, we can call it a fixed mixture design.

I have synthesized catalyst on my own and named them as C1, C2, C3, C4. Each catalyst has different catalytic power and are added in the same constant concentration.. Concentration of catalyst is not the objective here.
 
Last edited:
#5
Maybe it turns out that the 15 raw material consists of factors like a 2*2*3 + 3 additional, så that the 15 are created partly as a factorial. Then it can be formulated as a fractional factorial.

Tell us more.
Thanking you for the reply
I have 3 factors ( categorical) : factor 1(raw material) has 15 levels, factor 2(supplement) has 8 levels and factor 3(catalyst) has 4 levels.